#five steps

1 tagged report

Agent Zero Lets You Approve Each Action

Agent Zero 2.15, released 1 November 2026, changes how you decide what your AI assistant is allowed to do. Instead of one yes-or-no switch per tool, you can now approve or block individual actions inside each tool: read my files, yes; rewrite them, no. Every action gets three positions — On, Default or Off — so new tools follow your house rule automatically instead of arriving pre-approved. The release also closes two quiet loopholes. Telegram and WhatsApp senders can no longer lift their own restrictions; settings changes now require explicit administrators. And the remote typing tool now needs the same permission as the remote terminal, so blocking one really blocks both. Think of it as moving from handing over the whole key ring to cutting keys room by room. The same question applies to every app in your life: what can it do without asking you first, and who else can tell it what to do? #AgentZero #AIAgents #LeastPrivilege #AISecurity #OWASP #Privacy #OpenSourceAI — Links — https://agent-zero.ai/ https://github.com/agent0ai/agent-zero https://genai.owasp.org/llmrisk/llm062025-excessive-agency/ https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/ https://csrc.nist.gov/glossary/term/least_privilege a dumb drop by dumbfoundry

2026-11-02
All Reports →
dumb

dumb is the unspoken

Learn to build freely!